Metraipolis › archive › dc_2e786ea4d465

Efficient Auto-Interpretability of AI Models in Biology — arXiv:2608.27754

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.27754
License (as archived): CC0-1.0
Archived: 2026-10-17T10:47:55.495Z
SHA-256 of the archived text: 63f654c9757033e052aae35ec3c52f04620b4bd091a30b6a9c5bda0237df4e75
Stable URL of this page: https://metraipolis.com/documento/dc_2e786ea4d465

How to cite

Efficient Auto-Interpretability of AI Models in Biology — arXiv:2608.27754 — arXiv (en), https://arxiv.org/abs/2608.27754 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2e786ea4d465 (Metraipolis archive dc_2e786ea4d465, sha256 63f654c9757033e052aae35ec3c52f04620b4bd091a30b6a9c5bda0237df4e75)

BibTeX

@misc{dc2e786ea4d465,
  title = {Efficient Auto-Interpretability of AI Models in Biology — arXiv:2608.27754},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.27754},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2e786ea4d465; sha256 63f654c9757033e052aae35ec3c52f04620b4bd091a30b6a9c5bda0237df4e75},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27754. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Sparse autoencoders (SAEs), and other interpretability methods could turn AI models in Biology and other fields into engines of scientific discovery by explaining the superhuman capabilities of those models. However, a latent is only useful if we know three things: whether it is coherent, whether it can be described, and whether that description has predictive power. These questions are routinely conflated. We assemble them into a single pipeline and report the practical innovations each stage required. First, cross-seed dictionary stability prioritises which latents are worth spending resources to investigate. Second, an intruder-detection task asks whether a latents activating examples share a recognizable pattern. Third, a separate pass proposes a candidate biological description which we convert into falsifiable predictions which can be tested in silico. Deployed on the Boltz-1 Pairformer trunk, stability prioritisation finds interpretable latents using about 4.4 times fewer latent evaluations each, and at 5.2 times lower measured cost, while recovering over half of them, and the external check shows the surfaced motifs are significantly enriched for their claimed annotations. The results also suggest a possible tension: the cross- seed stability might be selecting for some types of features, like structure-related ones, much more than others, such as function-related features.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.AI).
Categorie arXiv: Quantitative Methods (q-bio.QM) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.27754 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T10:33:42.469Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).